Search overview

The search takeover replaces Shopify's native search with queries against your own index. Same engine as the collection grid, same fitment awareness, same speed.

How a query resolves#

1. Normalize The raw query is lowercased, trimmed, and run through your synonym rules — so "chevy" finds Chevrolet and "cat back" finds cat-back.
2. Look for a SKU An exact SKU match short-circuits everything and redirects straight to that product page. Staff and repeat buyers type SKUs; making them read a one-result page is a small insult.
3. Look for a vehicle Tokens are tested against the taxonomy. "f150 6 inch lift kit" yields the F-150 plus the remaining terms.
4. Search The leftover terms run against title, vendor, SKU, product type and your spec fields, typo-tolerant, ranked by the ranking tuner.
5. Filter If a vehicle is active — whether pre-existing or detected in step 3 — fitment filtering applies unless the shopper turns it off.

Vehicle detection#

Detection is conservative. It requires enough tokens to identify one vehicle unambiguously; "ford" alone sets nothing, because narrowing a search to every Ford ever made is not helpful.

When a vehicle is detected you can either set it as active automatically, or show a confirm chip ("Searching for your 2019 F-150 — not your vehicle?"). The chip is slightly slower and considerably less annoying; it is the default.

What's searchable#

Caching#

Popular queries are served from Redis, keyed by shop, query, vehicle and filters. That is why the tenth shopper searching "f150 lift kit" this hour gets an instant response — and why analytics is recorded at route level rather than inside the cache layer, so cached hits still count.